Mojo module
block_scaled_output_writer
Output tile writer for block-scaled matmul epilogue.
Implements TMEM → Registers → SMEM → GMEM pipeline using TMA stores. Parameterized on BlockScaledMatmulConfig instead of standard MatmulConfig.
Structs
-
BlockScaledTileWriter: Epilogue writer: TMEM → Registers → SMEM → GMEM via TMA stores.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!